From dc801c07cf38b0c495686463e6ca6f871a64440e Mon Sep 17 00:00:00 2001
From: YurenHao0426 <blackhao0426@gmail.com>
Date: Tue, 27 Jan 2026 09:57:37 -0600
Subject: Add collaborativeagents module and update gitignore

- Add collaborativeagents subproject with adapters, agents, and evaluation modules
- Update .gitignore to exclude large binary files (.whl, .tar), wandb logs, and results

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
---
 .../slurm/fullscale/test_25parallel.sh             | 66 ++++++++++++++++++++++
 1 file changed, 66 insertions(+)
 create mode 100644 collaborativeagents/slurm/fullscale/test_25parallel.sh

(limited to 'collaborativeagents/slurm/fullscale/test_25parallel.sh')

diff --git a/collaborativeagents/slurm/fullscale/test_25parallel.sh b/collaborativeagents/slurm/fullscale/test_25parallel.sh
new file mode 100644
index 0000000..09d5ddb
--- /dev/null
+++ b/collaborativeagents/slurm/fullscale/test_25parallel.sh
@@ -0,0 +1,66 @@
+#!/bin/bash
+#SBATCH --job-name=test_25parallel
+#SBATCH --account=bfqt-delta-gpu
+#SBATCH --partition=gpuA100x4
+#SBATCH --nodes=1
+#SBATCH --ntasks=1
+#SBATCH --cpus-per-task=16
+#SBATCH --gres=gpu:nvidia_a100:2
+#SBATCH --mem=128G
+#SBATCH --time=1:00:00
+#SBATCH --output=test_25parallel_%j.out
+#SBATCH --error=test_25parallel_%j.err
+
+# Quick test: 25 profiles × 2 sessions × 1 method (vanilla) = 50 sessions
+# With 25 parallel profiles to measure realistic throughput
+
+cd /projects/bfqt/users/yurenh2/ml-projects/personalization-user-model
+source /u/yurenh2/miniforge3/etc/profile.d/conda.sh
+conda activate eval
+
+export HF_HOME=/projects/bfqt/users/yurenh2/hf_cache/huggingface
+export PYTHONPATH="${PWD}/src:${PWD}/collaborativeagents:${PYTHONPATH}"
+export NCCL_P2P_DISABLE=1
+
+set -a
+source .env
+set +a
+
+pip install --quiet openai python-dotenv json-repair
+
+MODEL_8B="/projects/bfqt/users/yurenh2/ml-projects/personalization-user-model/models/llama-3.1-8b-instruct"
+
+pkill -f "vllm.entrypoints" 2>/dev/null || true
+sleep 2
+
+CUDA_VISIBLE_DEVICES=0 python -m vllm.entrypoints.openai.api_server \
+    --model $MODEL_8B --port 8003 --tensor-parallel-size 1 \
+    --gpu-memory-utilization 0.5 --max-model-len 8192 \
+    --dtype bfloat16 --disable-log-requests &
+
+for i in $(seq 1 90); do
+    curl -s http://localhost:8003/health > /dev/null 2>&1 && break
+    sleep 2
+done
+echo "vLLM ready."
+
+cd collaborativeagents/scripts
+
+# Test with 25 parallel profiles
+python run_experiments.py \
+    --methods vanilla \
+    --datasets math-hard \
+    --n-profiles 25 \
+    --n-sessions 2 \
+    --max-turns 8 \
+    --use-vllm \
+    --use-openai-user \
+    --openai-user-model gpt-5-mini \
+    --reward-mode llm \
+    --vllm-agent-url http://localhost:8003/v1 \
+    --parallel-profiles 25 \
+    --profile-path ../data/complex_profiles_v2/profiles_200.jsonl \
+    --output-dir ../results/test_25parallel_$(date +%Y%m%d_%H%M%S)
+
+echo "Test complete!"
+pkill -f "vllm.entrypoints" 2>/dev/null || true
-- 
cgit v1.2.3